Nano Banana 2 Lite

gemini-3.1-flash-lite-image (officially Nano Banana Lite, also referred to as Nano Banana 2 Lite in launch materials) is Google's image generation and editing model, released to general availability in June 2026. Built on the Gemini 3.1 Flash Lite architecture with a knowledge cutoff of January 2025, it is positioned as the "efficiency specialist" of the image generation family — below the generalist Nano Banana 2 (gemini-3.1-flash-image) and the high-fidelity Nano Banana Pro (gemini-3-pro-image). Google recommends it as the migration target for the legacy Nano Banana (gemini-2.5-flash-image).

Its core breakthrough is high-throughput generation at significantly reduced TPU compute cost: Google targets a sub-2 second end-to-end latency, enabling high-volume interactive developer use cases and real-time consumer applications. While prioritizing speed and cost efficiency, it maintains character alignment matching the original Nano Banana standard and legible in-image text rendering, with native support for interleaved generation and editing.

Key Capabilities

• Sub-2s Latency: Targets sub-2 second end-to-end latency, suited to interactive development and real-time consumer applications.
• Interleaved Generation & Editing: Natively supports Text → Text + Image(s) and Image + Text → Text + Image(s), returning text and imagery in one turn.
• Multi-turn Local Edits: Enables fast local changes such as color swaps, sticker creation, and background adjustments for iterative refinement.
• Character Alignment: Maintains character consistency matching the original Nano Banana standard, supporting storyboarding and virtual try-on use cases.
• Aspect Ratios: Supports 14 discrete aspect ratios (including 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9) spanning standard, portrait, and widescreen formats.
• 1K Optimization: Optimized for 1024×1024 (1K); the only supported image_size is 1024px (2K and 4K are unsupported).
• Provenance Watermarking: Always-on SynthID invisible watermark plus C2PA content credentials.

Technical Strengths

Feature	Benefit
Ultra-low-latency architecture	Sub-2s end-to-end latency target enables near-real-time, high-concurrency interactive and consumer apps with minimal wait friction
Cost efficiency	Significantly reduced TPU compute cost sharply lowers the expense of high-frequency generation (launch materials cite ~$0.034 per 1K image)
1K-resolution focus	Fixed 1024px (1K) output trades higher resolution for speed and cost, fitting high-throughput pipelines
Interleaved I/O	Native interleaved text + image input and output lets one turn cover "describe → generate → explain"
Provenance & compliance	SynthID (always on) + C2PA watermarking makes AI-generated content identifiable and compliance-friendly
Integration-friendly	Batch API and function calling support integration into automated asset pipelines and agentic workflows

Use Cases

• High-concurrency consumer apps: Powers real-time image generation for social and creative apps serving large user bases.
• Ad & marketing A/B testing: Rapidly generates ad variations and localized versions to accelerate creative validation.
• Interactive prototyping: Iterates from blank page to design concept at very low latency for fast visual exploration.
• E-commerce & virtual try-on: Uses character/object consistency to build try-on and product-display assets.
• Storyboarding & content creation: Maintains character consistency across generations to power storyboarding tools.
• Batch asset automation: Generates images at scale via the Batch API for programmatic content pipelines.
• Multi-turn refinement: Polishes a single asset step by step via local edits like color swaps, stickers, and background changes.

FAQ

What is gemini-3.1-flash-lite-image (Nano Banana Lite)?

It is the efficiency-tier model of Google's image generation family, officially named Nano Banana Lite, built for ultra-low-latency, low-cost image generation and editing. Its API code is gemini-3.1-flash-lite-image.

How does it differ from Nano Banana 2 and Nano Banana Pro?

All three belong to the Nano Banana family — Lite prioritizes speed and cost, Nano Banana 2 (gemini-3.1-flash-image) is the generalist, and Nano Banana Pro (gemini-3-pro-image) targets high fidelity and complex reasoning.

How fast is it and how much does it cost?

Google targets a sub-2 second end-to-end latency; launch materials separately cite ~4-second text-to-image generation (default thinking level) at roughly $0.034 per 1K image.

What resolutions and aspect ratios does it support?

It outputs only 1024px (1K) and supports 14 discrete aspect ratios (including 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9); 2K and 4K are unsupported.

Does it support image editing, and are outputs watermarked?

Yes — it supports interleaved editing and multi-turn local edits (color swaps, stickers, background adjustments); every output carries an always-on SynthID invisible watermark plus C2PA content credentials.

Should I migrate from the legacy Nano Banana?

Google recommends that users of the legacy Nano Banana (gemini-2.5-flash-image) migrate to this model for faster speed and lower cost.